Sam Apr 2026

The first paper, "Segment Anything," introduced the concept of a promptable segmentation system and the massive SA-1B dataset.

The most recent advancement, "SAM 3: Segment Anything with Concepts," unifies detection, segmentation, and tracking with text and visual prompts. Specialized "SAM" Papers The first paper, "Segment Anything," introduced the concept

Depending on your field, you might be looking for these "proper" adaptations: The first paper

The follow-up paper, "SAM 2: Segment Anything in Images and Videos," extends the model to handle video objects in real-time. "SAM 3: Segment Anything with Concepts